“The Nobel Prize in Physiology or Medicine was shared in 1979 by a physicist, Allan Cormack, and an engineer, Godfrey Hounsfield…. Hounsfield, working independently [of Cormack], built the first clinical [computed tomography] machine, which was installed in 1971. It was described in 1973 in the British Journal of Radiology. The Nobel Prize acceptance speeches (Cormack 1980; Hounsfield 1980) are interesting to read. A neurologist, William Oldendorf, had been working independently on the problem but did not share in the Nobel Prize…”
Oddly, Russ and I did not include Hounsfield’s 1973 paper in our list of references. I decided to dig it up and have a look. The reference and abstract is:
Hounsfield GN (1973) Computerized transverse axial scanning (tomography): Part I. Description of the system. Br J Radiol 46:1016–1022
This article describes a technique in which X-ray transmission readings are taken through the head at a multitude of angles: from these data, absorption values of the material contained within the head are calculated on a computer and presented as a series of pictures of slices of the cranium. The system is approximately 100 times more sensitive than conventional X-ray systems to such an extent that variations in soft tissues of nearly similar density can be displayed.
A dozen comments:
- This is Hounsfield’s most highly cited paper, with 4667 citations according to Google Scholar. That’s a respectable number (ten times more than any of my papers have), yet seems curiously small for a Nobel Prize-winning advance.
- Hounsfield’s paper is the first of a trilogy. Hounsfield is not a coauthor on the other two; they report clinical studies using the new technique.
- Hounsfield lists his institution as “Central Research Laboratories of EMI Limited”. EMI is famous in the music industry; it is the recording label responsible for the early hits of the Beatles.
- Hounsfield’s paper has only three references: two to his own preliminary reports and one to an article by Oldendorf. He didn’t cite Cormack’s papers.
- Hounsfield sounds most impressed not by recreating three-dimensional images from two-dimensional projections (which to me is the big advance) but instead by the increased sensitivity of the technique to small differences in x-ray absorption coefficient.
- Figure 3, illustrating the scanning device and sequence, is similar to Fig. 16.25 in IPMB.
- Hounsfield measured 160 points in each translation and performed 180 rotations. Each two-dimensional image was represented by an 80 × 80 grid of pixels.
- The reconstruction method was different from the two Russ and I analyze in Chapter 12 of IPMB: i) Fourier Transform Reconstruction and ii) Filtered Back-Projection. Instead, Hounsfield just fit his data using the least squares method (see Section 11.1 of IPMB). Hounsfield writes “Each beam path [in the CT scan], therefore, forms one of a series of 28,800 simultaneous equations, in which there are 6,400 variables and, providing that there are more equations than variables, the values of each [pixel] …. can be solved.”
- The Hounsfield unit was introduced in Fig. 9, but he did not, of course, call it that. Interestingly, his definition is different than what is used today. Equation 16.25 in IPMB gives the Hounsfield unit as H=1000(μtissue -μwater)/μwater), where μtissue and μwater are x-ray attenuation coefficients. In his paper, Hounsfield defines the unit the same way, except he replaces 1000 by 500.
- The article describes preliminary experiments using an iodine-containing contrast agent and digital subtraction, analogous to Fig. 16.23 in IPMB.
- The computer equipment pictured in Hounsfield’s paper look big and clunky today. I can only guess what paltry computer power he had available for these first reconstructions.
- I love the British Journal of Radiology, known as BJR. [What journal did you think that Bradley John Roth would like?]
I’ll conclude with Hounsfield’s final paragraph. To my ear, it sounds like classic British understatement.
It is possible that this technique may open up a new chapter in X-ray diagnosis. Previously, various tissues could only be distinguished from one another if they differed appreciably in density. In this procedure absolute values of the absorption coefficient of the tissues are obtained. The increased sensitivity of computerized X-ray section scanning thus enables tissues of similar density to be separated and a picture of the soft tissue structure within the cranium to be built up.